Skip to content
SortVisionAI Control Platform

The control software already running on 1,140+ sortation systems — flagging jams 14 seconds before fault.

SortVisionAI is the proprietary control layer behind every Sortons cross-belt, tilt-tray, shoe, and bombay sorter shipped since 2011. One platform. One telematics cloud. One SLA-backed uptime contract — 99.94% average across the trailing 24 months.

Book a Sortation Assessment View System Family →
Systems under live monitoring
1,140+
Average uptime, trailing 24 months
99.94%
Jam-detection lead time
14 sbefore fault
Deployed sort points, 38 countries
4,200+
Line-art schematic of a Sortons sortation loop showing throughput, routing accuracy, and uptime callouts.
SCH-01 SortVisionAI loop reference architecture
Reference Architecture / Tier 0–3

From sensor edge to the SortVisionAI Cloud — four tiers, one deterministic control plane.

SortVisionAI is not a black box wrapped around a third-party WCS. Every tier is owned, built, and supported in-house by our 89 controls engineers and 34 software developers — zero subcontracted commissioning. The result is a stack a controls engineer can review end-to-end before signing the PO.

Tier 0 · Sensor Edge

Field I/O & Vision Nodes

IP67-rated photoeyes, encoder arrays, and on-sorter GigE vision cameras stream parcel geometry at 120 Hz. Edge inference runs a quantized CNN classifier (parcel class & orientation) with sub-8 ms latency. No raw video leaves the sorter — only labeled events and 64-bit feature vectors.

  • 120 Hz vision sampling, <8 ms edge inference
  • Ethernet/IP, PROFINET, IO-Link fieldbus
  • No raw imagery egress — events only
Tier 1 · Controls

PLC & Safety Controllers

Rockwell ControlLogix and Siemens S7-1500 PLCs execute the real-time sort logic on a SIL 3-rated safety core. SortVisionAI's deterministic dispatcher runs the conveyor, diverter, and recirculation logic with a 2 ms scan budget — verified by an independent MHI Performance Certification in 2024.

  • SIL 3 safety core, 2 ms scan budget
  • Rockwell & Siemens native, no middleware shim
  • Hot-standby redundant controller pair
Tier 2 · On-Prem Server

SortVisionAI Core · Operator HMI

An industrial server on the customer network runs the SortVisionAI Core: anomaly-detection model, digital-twin simulator, and the WMS/WCS adapter layer. Verified integrations today include Manhattan Active WM, Blue Yonder WCS, SAP EWM, and Körber WMS. Operator HMI exposes live throughput, lane balance, and exception queues in a single pane.

  • Customer-site deployment, air-gapped option
  • Verified WMS/WCS adapters listed above
  • Single-pane HMI: throughput, lanes, exceptions
Tier 3 · Cloud

SortVisionAI Cloud · Telemetry

Anonymized telemetry streams over a mutually authenticated TLS 1.3 channel to the SortVisionAI Cloud — the same SOC 2 Type II environment that monitors 1,140+ active systems today. Predictive models retrain nightly on the federated fleet; field engineers receive pushed work-orders before the customer's shift starts.

  • SOC 2 Type II, TLS 1.3, mTLS device identity
  • Federated retraining across 1,140+ systems
  • Pushed work-orders to 140+ field engineers
Capability Modules

Three predictive modules, each named with the model class that does the work.

We avoid the phrase "AI-powered" without a model class behind it. Below is exactly what runs inside SortVisionAI today, what it measures, and what it lets your operations team do at 03:00 when a lane starts drifting.

M-01

Predictive Jam-Detection

Model class: multivariate time-series anomaly detection (Isolation Forest + LSTM autoencoder ensemble)

Each sort point streams 47 features into a per-station anomaly model — motor current ripple, belt slip, parcel inter-arrival time, and pneumatic pressure. When the ensemble crosses its calibrated threshold, the operator HMI shows a yellow pre-fault on the affected lane an average of 14 seconds before a hard jam. Median 11-week install means the model is training on real parcel data before most competitors have finished their kickoff meeting.

  • 47-feature per-station telemetry, 1 kHz
  • 14 s median pre-fault lead time
  • Auto-thresholds per SKU mix, no manual tuning
SortVisionAI predictive jam-detection HMI panel showing a 14-second pre-fault alert on a single sort station.
M-02

Vision-Based Routing

Model class: computer vision — YOLOv8-nano detector + transformer-based parcel embedding

On-sorter GigE cameras detect parcels at induction, classify them against your SKU profile, and write the routing decision to the PLC in under 30 ms. The transformer embedding lets the system recognize unfamiliar parcel geometries it has never seen in training — useful during peak when a 3PL adds a new customer mid-season. Verified routing accuracy across 4,200+ deployed sort points: 99.98%.

  • YOLOv8-nano detector, 30 ms decision budget
  • Transformer embedding handles novel SKU profiles
  • 99.98% routing accuracy, 4,200+ sort points
Schematic pipeline diagram of the SortVisionAI vision-based routing module from induction camera to PLC write.
M-03

Throughput Optimization

Model class: digital-twin simulation + constrained linear-programming lane balancer

SortVisionAI keeps a continuously updated digital twin of your sorter — geometry, lane capacities, induction cadence, current SKU mix. Every 30 seconds the optimizer re-solves a constrained lane-balancing LP and pushes new setpoints to the PLC. The result is sustained peak throughput of 24,000 items per hour across cross-belt and tilt-tray configurations, with MHI-verified certification on record.

  • Live digital twin, 30 s re-solve cadence
  • Constrained LP lane balancer
  • 24,000 items/hr peak — MHI certified, 2024
SortVisionAI throughput optimization dashboard showing a live digital twin, balanced lane loads, and an LP re-solve log.
SORTVISIONAI · LIVE OPERATIONS READOUT STREAMING · 02:47 UTC · NORTH AMERICA + EMEA

Measured right now across 1,140+ active sortation systems.

1140+
Active sortation systems under live monitoring
SortVisionAI Cloud · Telemetry ping 5s
99.94%
Trailing 24-month uptime across the installed base
Contractual SLA tier available up to 99.97%
14s
Median fault-prediction lead time before a jam event
Predictive maintenance · SortVisionAI v4.2
38 ◐
Countries with deployed sort points
4,200+ live sort destinations

Telemetry is sampled at the PLC layer and streamed to the SortVisionAI control plane. Numbers shown are the live aggregate of the Sortons installed base — not a marketing claim.

ENTERPRISE POSTURE · ANSWER FIRST

Trailing 99.94% uptime. Contractual tiers up to 99.97%. Built for Fortune 1000 procurement review.

Sortons operates under ISO 9001:2015 and aligns to ISO 27001 controls for information security management. The SortVisionAI cloud is monitored from a NOC staffed by Sortons engineers in Atlanta and Eindhoven — no third-party operators handle your parcel telemetry.

01

Uptime SLA tiers

  • Standard99.90% — site-level credit ladder
  • Trailing99.94% measured across the installed base, last 24 months
  • Premium99.97% contractual, with 4-hour on-site response in 14 U.S. metros
02

Security & compliance controls

  • ISO 27001Aligned ISMS, annual third-party audit
  • EncryptionTLS 1.3 in transit · AES-256 at rest
  • AccessSSO/SAML, role-based access, hardware-MFA for engineering
  • AuditImmutable event log retained 13 months
  • Plant safetyOSHA VPP Star worksite since 2021 · CE / UL certified
03

Data residency

Customer telemetry is stored in the region it is generated. North American facilities land in our Virginia and Oregon data centers; EMEA facilities in Frankfurt and Dublin. Customer-managed keys are available on the Premium tier.

04

OTA release cadence

  • Security patchesWithin 72 hours of CVE triage
  • Minor releasesBi-weekly, signed and rollback-capable
  • Major releasesQuarterly, with a 30-day customer opt-in window
  • Change windowsCustomer-defined; no surprise reboots

Need our SOC 2 letter, a customer reference, or a network diagram? They ship with every sortation assessment.

Download the posture brief →
FROM THE SORT FLOOR · NAMED ENGAGEMENT

We routed the 2022 holiday peak — 4.1 million parcels a day through Sortons sortation — without a single unplanned stop on the cross-belt line. SortVisionAI flagged two jams before they happened, and the Sortons engineer was already on-site to clear them.

M. Halverson VP, Fulfillment Operations DHL Supply Chain — North American e-commerce hub
Engagement Cross-belt sortation, 24,000 items/hr peak Live since 2022